Numerical Methods for Differential Equations
Contents
• Review of numerical integration methods
– Rectangular Rule
– Trapezoidal Rule
– Simpson’s Rule
• How to make a connect-the-dots graphic
• Numerical Methods for y 0 = F (x)
– Maple code for Rect, Trap, Simp methods
• Numerical Methods for y 0 = f (x, y)
– Maple code for Euler, Heun, RK4 methods
• Methods for planar systems
• Methods for n × n systems
y
Rectangular Rule. The approximation uses Euler’s idea of replacing the integrand by a constant.
The value of the integral is approximately the area of a rectangle of width b − a and height F (a).
Z
F
x
b
a
b
F (x)dx ≈ (b − a)F (a)
a
y
Trapezoidal Rule. The rule replaces the integrand F (x) by a linear function L(x) which connects the planar points (a, F (a)), (b, F (b)). The value of the integral is approximately the area
under the curve L, which is the area of a trapezoid.
Z
b
F (x)dx ≈
b−a
2
a
x
a
Simpson’s Rule. The rule replaces the integrand F (x) by a quadratic polynomial Q(x) which
connects the planar points (a, F (a)), ((a + b)/2, F ((a + b)/2)), (b, F (b)). Then the integral
of F is approximately the area under the quadratic curve Q.
b
a

F (x)dx ≈ (b − a) 
b
(F (a) + F (b))
y
Z
F
L
F (a) + 4F
a+b
2
6
+ F (b)


Q
F
x
a
b
How to make a connect-the-dots graphic
x
0.0
0.1
0.2
0.3
0.4
0.5
y
2.000
1.901
1.808
1.727
1.664
1.625
x
0.6
0.7
0.8
0.9
1.0
y
1.616
1.643
1.712
1.829
2.000
y
x
The table consists of xy -values for y = x3 − x + 2 . The graphic represents the table’s
rows, which are pairs (x, y), as dots. Joined dots make the connect-the-dots graphic.
Maple code
A connect-the-dots graphic can be made in maple by supplying a list L of pairs to be
connected. An example:
L:=[[1,3],[2,1],[3,5]]:
plot(L);
Numerical Methods for y 0 = F (x)
Rx
Quadrature applies to give y(x) = y0 + x0 F (x)dx. Numerical solution methods
amount to approximating the integral on the right by Rectangular, Trapezoidal and Simpson
methods.
R x0+h
The methods replace the exact value of an integral x0
F (x)dx by a numerical approximation value which is useful for graphics when h is small. Larger intervals are broken into
smaller intervals of length h, then the approximation is applied.
Table 1. Three numerical integration methods.
Rect Y = y0 + hF (x0 )
Trap Y = y0 +
Simp Y = y0 +
h
2
h
6
(F (x0) + F (x0 + h))
(F (x0) + 4F (x0 + h/2) + F (x0 + h)))
Maple code for the Rectangular and Trapezoid Rules
# Rectangular algorithm
# Trapezoidal algorithm
# Group 1, initialize.
# Group 1, initialize.
F:=x->evalf(cos(x) + 2*x): F:=x->evalf(cos(x) + 2*x):
x0:=0:y0:=0:h:=0.1*Pi:
x0:=0:y0:=0:h:=0.1*Pi:
Dots1:=[x0,y0]:
Dots2:=[x0,y0]:
# Group 2, repeat 10 times
Y:=y0+h*F(x0):
x0:=x0+h:y0:=evalf(Y):
Dots1:=Dots1,[x0,y0];
# Group 2, repeat 10 times
Y:=y0+h*(F(x0)+F(x0+h))/2:
x0:=x0+h:y0:=evalf(Y):
Dots2:=Dots2,[x0,y0];
# Group 3, plot.
plot([Dots1]);
# Group 3, plot.
plot([Dots2]);
Maple code for Rectangular and Simpson Rules
# Rectangular algorithm
# Simpson algorithm
# Group 1, initialize.
# Group 1, initialize.
F:=x->evalf(exp(-x*x)):
F:=x->evalf(exp(-x*x)):
x0:=0:y0:=0:h:=0.1:
x0:=0:y0:=0:h:=0.1:
Dots1:=[x0,y0]:
Dots3:=[x0,y0]:
# Group 2, repeat 10 times
Y:=evalf(y0+h*F(x0)):
x0:=x0+h:y0:=Y:
Dots1:=Dots1,[x0,y0];
# Group 2, repeat 10 times
Y:=evalf(y0+h*(F(x0)+
4*F(x0+h/2)+F(x0+h))/6):
x0:=x0+h:y0:=Y:
Dots3:=Dots3,[x0,y0];
# Group 3, plot.
plot([Dots1]);
# Group 3, plot.
plot([Dots3]);
Numerical Methods for y 0 = f (x, y)
The methods replace the exact value of
Z
x0 +h
y(x0 + h) = y0 +
f (x, y(x))dx
x0
by a numerical approximation Y . The value is useful for graphics when h is small.
Table 2. Three numerical methods for y 0 = f (x, y).
Euler Y = y0 + hf (x0 , y0 )
Heun
y1 = y0 + hf (x0, y0)
Y = y0 +
RK4
h
(f (x0, y0) + f (x0 + h, y1))
2
k1 = hf (x0, y0)
k2 = hf (x0 + h/2, y0 + k1/2)
k3 = hf (x0 + h/2, y0 + k2/2)
k4 = hf (x0 + h, y0 + k3)
k1 + 2k2 + 2k3 + k4
Y = y0 +
6
Maple code for Euler and Heun methods
# Euler algorithm
# Group 1, initialize.
f:=(x,y)->-y+1-x:
x0:=0:y0:=3:h:=0.1:L:=[x0,y0]:
# Heun algorithm
# Group 1, initialize.
f:=(x,y)->-y+1-x:
x0:=0:y0:=3:h:=0.1:L:=[x0,y0]:
# Group 2, repeat 10 times
Y:=y0+h*f(x0,y0):
x0:=x0+h:y0:=Y:L:=L,[x0,y0];
# Group 2, repeat 10 times
Y:=y0+h*f(x0,y0):
Y:=y0+h*(f(x0,y0)+f(x0+h,Y))/2:
x0:=x0+h:y0:=Y:L:=L,[x0,y0];
# Group 3, plot.
plot([L]);
# Group 3, plot.
plot([L]);
Maple code for Heun and RK4 methods
# Heun algorithm
# Group 1, initialize.
f:=(x,y)->-y+1-x:
x0:=0:y0:=3:h:=0.1:L:=[x0,y0]:
# RK4 algorithm
# Group 1, initialize.
f:=(x,y)->-y+1-x:
x0:=0:y0:=3:h:=0.1:L:=[x0,y0]:
# Group 2, repeat 10 times
Y:=y0+h*f(x0,y0):
Y:=y0+h*(f(x0,y0)+f(x0+h,Y))/2:
x0:=x0+h:y0:=Y:L:=L,[x0,y0];
# Group 2, repeat 10 times.
k1:=h*f(x0,y0):
k2:=h*f(x0+h/2,y0+k1/2):
k3:=h*f(x0+h/2,y0+k2/2):
k4:=h*f(x0+h,y0+k3):
Y:=y0+(k1+2*k2+2*k3+k4)/6:
x0:=x0+h:y0:=Y:L:=L,[x0,y0];
# Group 3, plot.
plot([L]);
# Group 3, plot.
plot([L]);
Numerical Algorithms: Planar Case
Notation. Let t0 , x0 , y0 denote the entries of the dot table on a particular line. Let h be
the increment for the dot table and let t0 + h, x, y stand for the dot table entries on the
next line.
Planar Euler Method.
x = x0 + hf (t0, x0, y0),
y = y0 + hg(t0, x0, y0).
Planar Heun Method.
x1
y1
x
y
=
=
=
=
x0 + hf (t0, x0, y0),
y0 + hg(t0, x0, y0),
x0 + h(f (t0, x0, y0) + f (t0 + h, x1, y1))/2
y0 + h(g(t0, x0, y0) + g(t0 + h, x1, y1))/2.
Planar RK4 Method.
k1
m1
k2
m2
k3
m3
k4
m4
x
y
=
=
=
=
=
=
=
=
hf (t0, x0, y0),
hg(t0, x0, y0),
hf (t0 + h/2, x0 + k1/2, y0 + m1/2),
hg(t0 + h/2, x0 + k1/2, y0 + m1/2),
hf (t0 + h/2, x0 + k2/2, y0 + m2/2),
hg(t0 + h/2, x0 + k2/2, y0 + m2/2),
hf (t0 + h, x0 + k3, y0 + m3),
hg(t0 + h, x0 + k3, y0 + m3),
1
= x0 + (k1 + 2k2 + 2k3 + k4) ,
6
1
= y0 + (m1 + 2m2 + 2m3 + m4) .
6
Numerical Algorithms: General Case
Consider a vector initial value problem
u0(t) = F(t, u(t)),
u(t0) = u0.
Vector Euler Method
u = u0 + hF(t0, u0)
Vector Heun Method
w = u0 + hF(t0, u0),
h
u = u0 + (F(t0, u0) + F(t0 + h, w))
2
Vector RK4 Method
k1
k1
k1
k1
=
=
=
=
hF(t0, u0),
hF(t0 + h/2, u0 + k1/2),
hF(t0 + h/2, u0 + k2/2),
hF(t0 + h, u0 + k3),
1
u = u0 + (k1 + 2k2 + 2k3 + k4) .
6